Welcome to AI Agent Flow

Building software is complex. It involves much more than just writing code—you need to architect systems, review PRs, write tests, and hunt down elusive bugs.

AI Agent Flow solves this by putting a specialized team of AI agents directly on your local machine.

How it works

When you run aiagentflow run "Build a login system", you aren't just sending a prompt to an LLM. You are kicking off a structured, multi-agent workflow:

1. The Architect breaks down the task into modules and writes technical specs.
2. The Coder writes the code.
3. The Reviewer checks the code for anti-patterns and vulnerabilities.
4. The Tester writes unit and integration tests.
5. The Judge runs npm test and sends errors back to the Coder if needed.

No cloud lock-in. No sending your proprietary code to a closed platform. It all runs on your machine, integrating directly with your local Git repository.

Get Started

Ready to experience the future of local-first engineering? Head over to our Documentation to learn how to configure your agents, or check out our GitHub Repository to star the project and join the community.

Install AI Agent Flow today:

npm install -g aiagentflow

Stay tuned for more updates on how to build custom agents and integrate local LLMs like Ollama into your workflows!